BNY Mellon is advancing its artificial intelligence strategy by integrating Google Cloud’s agentic AI technology, including the latest Gemini 3 model, into its internal AI platform, Eliza. Named after Alexander Hamilton’s wife, Eliza is designed to streamline complex, repetitive tasks across the bank’s operations. Sarthak Pattanaik, BNY’s chief data and AI officer, said the integration aims to make workflows faster and more efficient. One key example is client onboarding, a process traditionally bogged down by manual document collection, identity verification, risk checks, and data entry. With agentic AI, Eliza can break down such tasks into smaller steps, coordinate actions across systems, and deliver results in a more seamless, orchestrated manner. Gemini 3, launched in mid-November, stands out for its ability to analyze text, images, tables, PDFs, and audio simultaneously—making it ideal for handling the diverse financial documents common in banking. This multimodal capability allows employees to upload mixed-format materials and receive synthesized insights quickly. BNY’s AI journey began gaining momentum in 2023. Since then, Eliza has been deployed across more than 120 automated tasks. Nearly all employees have completed mandatory training in generative AI and responsible AI practices. The bank previously partnered with OpenAI and invested in an AI supercomputer powered by NVIDIA, positioning itself as a pioneer among major financial institutions. Now, the integration with Google Cloud marks another milestone in its push to leverage cutting-edge AI. Safety and compliance remain top priorities. Both BNY and Google stress that agentic AI in a regulated environment must operate within strict boundaries. Each AI agent undergoes a rigorous internal model-risk review before deployment. Access controls limit what data agents can view or act on, and performance is monitored daily, feeding back into continuous improvement. Rohit Bhat, Google Cloud’s head of financial services, explained that agents are governed by development kits and communication protocols that define what they can do and with whom they can interact. These protocols ensure agents only communicate for specific, approved purposes and cannot exceed their designated roles. Bhat noted that financial institutions are ideal testing grounds for agentic AI due to their complex, rule-driven workflows and high compliance standards. Google’s goal is to ensure these systems operate not just intelligently, but responsibly—staying aligned with business policies, internal rules, and regulatory requirements. As major banks like Goldman Sachs, Morgan Stanley, and JPMorgan experiment with AI, BNY’s move highlights a growing trend: the shift from isolated AI tools to integrated, autonomous systems that can handle end-to-end processes. The success of this model will depend on balancing innovation with the strict safety and governance needed in the financial world.